Himanshu Raj

High performance and scalable I/O virtualization via self-virtualized devices

While industry is making rapid advances in system virtualization, for server consolidation and for improving system maintenance and management, it has not yet become clear how virtualization can contribute to the performance of high end systems. In this context, this paper addresses a key issue in system virtualization - how to efficiently virtualize I/O subsystems and peripheral devices. We have developed a novel approach to I/O virtualization, termed self-virtualized devices, which improves I/O performance by off loading select virtualization functionality onto the device. This permits guest virtual machines to more efficiently (i.e., with less overhead and reduced latency) interact with the virtualized device. The concrete instance of such a device developed and evaluated in this paper is a self-virtualized network interface (SV-NIC), targeting the high end NICs used in thehigh performance domain. The SV-NIC (1) provides virtual interfaces (VIFs) to guest virtual machines for an underlying physical device, the network interface, (2) manages the wayin which the devices physical resources are used by guest operating systems, and (3) provides high performance, low overhead network access to guest domains. Experimental results are attained in a prototyping environment using an IXP 2400-based ethernet board as a programmable network device. The SV-NIC scales to large numbers of VIFs and guests, and offers VIFs with 77% higher throughput and 53% less latency compared to the current standard virtualized device implementations on hyper visor-based platforms.

Mapping peer behavior to packet-level details: a framework for packet-level simulation of peer-to-peer systems

The growing interest in peer-to-peer systems (such as Gnutella) has inspired numerous research activities in this area. Although many demonstrations have been performed that show that the performance of a peer-to-peer system is highly dependent on the underlying network characteristics, much of the evaluation of peer-to-peer proposals has used simplified models that fail to include a detailed model of the underlying network. This can be largely attributed to the complexity in experimenting with a scalable peer-to-peer system simulator built on top of a scalable network simulator with packet-level details. In this work we design and develop a framework for an extensible and scalable peer-to-peer simulation environment that can be built on top of existing packet-level network simulators. The simulation environment is portable to different network simulators, which enables us to simulate a realistic large scale peer-to-peer system using existing parallelization techniques. We demonstrate the use of the simulator for some simple experiments that show how Gnutella system performance can be impacted by the network characteristics.

Bootstrapping in Gnutella: A Measurement Study

To join an unstructured peer-to-peer network like Gnutella, peers have to execute a bootstrapping function in which they discover other on-line peers and connect to them. Until this bootstrapping step is complete, a peer cannot participate in file sharing activities. Once completed, a peer’s search and download experience is strongly influenced by the choice of neighbor peers resulting from the bootstrapping step. Despite its importance, there has been very little attention devoted to understanding the behavior of this bootstrapping function. In this paper, we study the bootstrapping process of a peer in the Gnutella network. We find that (1) there is considerable variation among various servent implementations, and hence in their bootstrapping performance. (2) The neighbors of a peer, which are the outcome of the bootstrapping process, play a very important role in the peer’s search and download performance. (3) Even though the GWebCache system for locating peers is designed to operate as a truly distributed caching system, it actually operates more like a centralized infrastructure function, with significant load imbalance. (4) The GWebCache system is subject to significant misreporting of peer and GWebCache availability, due to stale data and absence of validity checks.

AutoPower: toward energy-aware software systems for distributed mobile robots

Autonomous robot systems have to manage their energy wisely in order to complete their missions. Typical approaches seek to conserve energy by energy-efficient motion or sensor planning. This paper puts forth a distributed systems approach to power management. Specifically, it develops and presents AutoPower, which is a model that characterizes robot software systems computation and communication energy behaviors. With AutoPower, it is possible to make principled decisions about (1) where to deploy software components across the distributed computing resources of autonomous robotic systems, and (2) how the different systems involved should communicate to best meet overall mission objectives. We showcase AutoPower by using a multi-robot search-and-rescue mission as a guiding application. For this scenario, application of the model shows that there are counter-intuitive energy trade-offs in configuring such application software. Further, by using AutoPower to guide deployment and interconnects at runtime, for certain configurations, overall computing system lifetimes can be increased by up to 57% over a base-line configuration

ContentCascade incremental content exchange between public displays and personal devices

Public exhibits or displays are widely used to present information in public and semipublic places. Users may pass by several of these and observe a significant amount of interesting content, much of which may be forgotten. Unfortunately, there is no widely adopted way for users to acquire related information in locale for their subsequent perusal. We present ContentCascade, a mechanism that allows users to implicitly download various levels of detail of summary information about the content available at the public displays to their personal devices. It also naturally and informally enables users to trade off quality of summary information with presence time. This way, ContentCascade can potentially enhance the overall user experience with public displays as users can recall the summary information later.

VMedia: enhanced multimedia services in virtualized systems

This paper presents the VMedia multimedia virtualization framework, for sharing media devices among multiple virtual machines (VMs). The framework provides logical media devices, exported via a well defined, higher level, multimedia access interface, to the applications and operating system running in a VM. By using semantically meaningful information, rather than low-level raw data, within the VMedia framework, efficient virtualization solutions can be created for physical devices shared by multiple VMs. Experimental results demonstrate that the base cost of virtual device access via VMedia is small compared to native physical device access, and in addition, that these costs scale well with an increasing number of guest VMs. Here, VMedias MediaGraph abstraction is a key contributor, since it also allows the framework to support dynamic restructuring, in order to adapt device accesses to changing requirements. Finally, VMedia permits platforms to offer new and enhanced logical device functionality at lower costs than those achievable with alternative solutions.

Extending virtualization services with trust guarantees via behavioral monitoring

Todays virtualized platforms enable virtualization services (VSs) that can offer enhanced functionalities to guest virtual machines (VMs) based on behavioral monitoring. One such set of functionality concerns protected service access, by having a VS impose access controls that can be altered and refined at runtime. Changes are made in accordance with the levels of trust associated with certain VMs - where VSs use runtime monitoring to derive current trust levels from observed guest VM behavior. This paper develops and evaluates implementation methods for trust enhancements of virtualization services and demonstrates their utility for a storage virtualization service, termed protected object store (POS). An implementation of POS based on the PVFS file system as a backend and using the Xen VMM as a virtualization infrastructure is shown effective in its ability to enforce fine-grained, role-based access controls on storage usage based on the VMs dynamic level of trust, while minimally impacting the overall performance of the storage service.

O2S2: enhanced object-based virtualized storage

Object based storage devices (OSDs) elevate the level of abstraction presented to clients, thereby permitting them to offer methods for managing, sharing, and securing information that go beyond those offered by block-based stores. The Object-Oriented Storage System (O2S2) architecture presented and evaluated in this paper provides object-based storage in a virtualized environment. This service provides a virtual object-based storage device (vOSD) to virtual machines. The use of vOSDs permits the service provider, i.e., the vOSD storage domain, to offer to guest virtual machines new methods for resource management and consolidation Methods demonstrated in this paper include improved support for access control and for heterogeneity of storage devices. A prototype PVFS-based O2S2 implementation demonstrates that its enhanced services can be provided at low cost, enabled in part by the efficient utilization of otherwise idle storage domain resources.

Spirits : using virtualization and pervasiveness to manage mobile robot software systems

Management capabilities comprise a key component of any autonomous distributed system. In this work our focus is on mobile systems like teams of robots exploring and operating in some physical environment. Here, basic management goals are to adapt application behavior to prevent or mitigate reductions in the applications quality of service. This paper presents the spirits system-level mechanisms supporting (1) behavior persistence – the ability to maintain some desirable behavior learned through online adaptation – and (2) behavior propagation – the ability to propagate a learned behavior across different physical components. With spirits, software management is enriched with a low-level mechanism, termed a spirit cache, which permits a mobile entity to cache its current state and code to realize behavior persistence. Next, a cached spirit can be acquired by a different physical component and then used, thereby propagating it. By using system-level virtualization techniques to realize spirit caching and propagation, both can be performed without the need to make any changes to application code, without requiring middleware-level support, and without changes to operating system kernels or utilities. Furthermore, any number of spirits can exist in a robot with system-level isolation guarantees. Experimental results presented in this paper highlight the types of overheads for spirit exchanges experienced on typical next generation virtualizable embedded machines, and indicate optimizations to be considered in future research.

Enabling semantic communications for virtual machines via iConnect

iConnect is an abstraction that encapsulates all of a virtual machines interactions with outside entities. Its intent is to exploit semantic information to better support the end-to-end requirements of such VM communications. Focusing on the I/O performed by VMs and leveraging the fact that modern systems already have to virtualize the physical devices used by VMs, this paper shows that the iConnect abstraction can be implemented with the extension of existing virtual device interfaces. Specifically, by devising enhanced virtual devices, we can (1) efficiently implement the communication paths between virtual machines (VMs) and the virtualized platforms (VPs) on which they run, and (2) capture semantic information about VM-device interactions, which can then be used to implement additional functionality and efficient sharing of physical devices. The paper presents three concrete realizations of the iConnect abstraction: a multimedia device virtualization solution (VMedia) which utilizes semantic information to implement efficient sharing and enhanced functionality, a network virtualization solution that provides virtual NICs with QoS-support where the VM communicates its QoS requirements to the VP, and a storage virtualization solution which permits a VM to access a block device regardless of whether such a device is physically located locally or must be accessed at a remote location. A Xen-based implementation of the iConnect concept demonstrates substantial performance improvements and additional functionality derived from their use at a minimal cost to VMs, in part because iConnect utilizes additional computational resources of the VP and can take better advantage of certain underlying platform capabilities.